Bayesian Polynomial Regression Models to Fit Multiple Genetic Models for Quantitative Traits.

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Genetic association studies commonly employ five distinct models: genotypic, additive, dominant, co-dominant, and recessive.
• Selecting the most appropriate genetic model is crucial for accurate association analysis.
• Existing methods may lack computational efficiency or comprehensive model comparison.

Purpose of the Study:

• To develop a unified Bayesian framework for selecting the most likely genetic model.
• To enhance computational efficiency by simultaneously fitting multiple genetic models.
• To provide a robust method for genetic model selection in association studies.

Main Methods:

• A polynomial parameterization of genetic data is utilized to fit five genetic models concurrently.
• A closed-form expression for the marginal likelihood is derived for normally distributed data.
• The proposed Bayesian framework is evaluated using both simulated and real genome-wide datasets.

Main Results:

• The proposed Bayesian framework offers a coherent and computationally efficient approach to genetic model selection.
• Simultaneous fitting of multiple models streamlines the analysis process.
• Performance evaluation demonstrates the method's efficacy on simulated and real-world genetic data.

Conclusions:

• The developed Bayesian framework provides a superior method for selecting the most appropriate genetic model in association studies.
• This approach enhances analytical efficiency and accuracy in genetic data interpretation.
• The method is broadly applicable to genome-wide association studies for improved biological insights.